1. A HYBRID FORECASTING MODEL BUILDING STRATEGY.
- Author
-
Aliyev, Vugar, Pkhovelishvili, Merab, and Archvadze, Natela
- Subjects
- *
FORECASTING , *SUFFICIENT statistics - Abstract
The article discusses the classification of existing models for predicting given events according to the statistics of all their forecasts and an algorithm for using the intersection or combination of sets of forecasts projections with other models to find the best pairs. The following terms were introduced: “necessary”, “sufficient”, “unnecessary” and “insufficient” forecasting models. This is discussed…this is discussed in the earthquake forecasting examples, and insufficient similarly be used to predict other events. We explained what “necessary” and “sufficient” models are. For “unnecessary” models, an algorithm is given for how to choose a hybrid model - the intersection of two or more models that together give a forecast with a higher probability. We also discussed “sufficient” models and an algorithm for selecting such “sufficient” models, the combination of which completely covers all past events, that is, the combination of such “sufficient” models becomes “necessary”. Discusses how one can obtain a "sufficient" or "nearly sufficient" forecasting model by combining "necessary" models, and by combining "sufficient" models to obtain a "necessary" or “almost necessary” models. Also unnecessary models is discussed. In our earlier work, these models were not taken into account; such models were removed from the model database. Similarly, when considering “sufficient” models, if the forecast of the occurrence of an event is specified redundantly, such a model can be excluded from the set of “sufficient” models. From “sufficient” models we obtain the “necessary” model, which will be both “sufficient” and “necessary” at the same time. In addition, we combine enough models to get the “necessary” model. All of this uses forecast statistics to strategically select a hybrid model. [ABSTRACT FROM AUTHOR]
- Published
- 2024